Neural network-based Smith predictor design for the time-delay in a tele-operated control system

This article presents a methodology for compensating for the time-delay effects in tele-operated control systems. Compensation can be carried out by a neural network. A tele-operated system consists of a master robot to give commands, and a slave robot to work with the environment. The positional command by the master robot is transferred to the slave robot, and the contact force from the environment is transferred back to the master robot. The structure of the Smith predictor is modified by replacing the linear estimator with a neural network whose structure is based on the radial basis function (RBF). The RBF network identifies the slave model to deal with the nonlinearities in the system. Simulation studies have been conducted, and experimental studies of one-directional force control were performed to confirm the simulation results.